YOLO-AMM: A Real-Time Classroom Behavior Detection Algorithm Based on Multi-Dimensional Feature Optimization.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Educational Technology

Background:

• Current classroom behavior detection models struggle with detailed feature capture, multi-layer feature correlation, and multi-scale target adaptability.
• Achieving high-precision, real-time detection in complex classroom environments remains a challenge.

Purpose of the Study:

• To propose an improved classroom behavior detection algorithm, YOLO-AMM, addressing limitations in existing models.
• To enhance the accuracy and real-time performance of intelligent classroom behavior analysis.

Main Methods:

• Developed the Adaptive Efficient Feature Fusion (AEFF) module to improve detailed feature capture and semantic information fusion.
• Designed the Multi-dimensional Feature Flow Network (MFFN) for enhanced multi-dimensional feature correlation using multi-scale aggregation and contextual diffusion.
• Introduced the Multi-Scale Perception and Fusion Detection Head (MSPF-Head) to adapt to various target scales through perception, interaction, and fusion mechanisms.

Main Results:

• YOLO-AMM demonstrated significant improvements over the YOLOv8n model, with a 3.1% increase in mAP0.5 and a 4.0% increase in mAP0.5-0.95.
• The algorithm achieved a detection speed of 169.1 frames per second (FPS), an increase of 12.9 FPS, meeting real-time detection requirements.

Conclusions:

• The proposed YOLO-AMM algorithm effectively addresses the limitations of existing models in classroom behavior detection.
• YOLO-AMM offers superior accuracy and real-time performance, making it suitable for intelligent educational environments.